Study On The Methods Of Source Number Estimation

Spatial spectrum estimation is a signal processing technology that developed in recent decades. It has widely applications in radar, communications, sonar and etc. The source number estimation is an important part of spatial spectrum estimation, and the most difficult problem to apply the spatial spectrum estimation theory into practice. Therefore, in the recent period, the spatial source number estimation became a hot research in this area. Many scholars had made a lot of algorithms both at home and abroad. Based on the needs of this application, this thesis discussed how to detect the number of space source signal.This dissertation is mainly studied on the approaches of the source number estimation. Firstly, taking MUSIC algorithm as an example, the effect of source number estimation to the estimation of DOA is discussed in this thesis, and their effects of the over- and under-estimation of source number to the MUSIC algorithm are analyzed in this thesis. And then, this dissertation is studied on some approaches of source number estimation in turn, for instance, the AIC and MDL methods based on the information theoretic criteria, Gerschgorin disk method, the method based on the spatial smoothing and Gerschgorin disk and the method based on the reverse conjugate array and Gerschgorin disk. This dissertation introduces the above methods in detail, according to their reasons, background and the specific principle of estimation. Then it judges of the above algorithms and simulates them to illuminate the estimations. Finally, this dissertation has emphatically studied on the method based on diagonal loading. It is proposed for the decreasing performance of the information theoretic method because of the array error and the environment noise in practice. Theoretical analysis and statistic analysis through computer simulations about this method have been done. We improve that the diagonal loading improved performance of AIC and MDL. In the final part, the striving direction of the source number estimation has been proposed.